Force Amplitude Control Algorithm on Fatigue Test with Sine Waveform

Transcription

1 Force Amplitude Control Algorithm on Fatigue Test with Sine Waveform Design Team Ryan Kenny, Phil Salisbury Nick Johnson, John Martin Design Advisor Prof. Andrew Gouldstone Abstract This report investigates the inner workings of a compression materials testing machine provided by ADMET. The system is ideal for fatigue life tests which are critical in mechanical designs. The system has been deemed faulty because it cannot reliably generate a cyclical force within the specified tolerance. Any sinusoidal waveform executed by the machine tends to result in an overshoot or undershoot from the force amplitude. It is suggested that the controls be scrutinized and adjusted to correct these errors. By investigating the sources that command the motor to adjust its actions, a solution may be found in programming. This paper proposes that a comprehensive understanding of controls and computation languages is needed to find a solution. It is suggested that the system be modeled in a more controllable environment so that the variations may be easier to find. The proportional-integral-derivative controller in particular is suspect to be reacting to the error in an incorrect manner. Other possible solutions mentioned are individual stiffness algorithms and increasing the linear actuator strength. Progress in exploring possible solutions or methods requires analyzing the relationship between the linear motor and the controller. The structural integrity is also under scrutiny for mechanical analysis separate from possible control errors. This report looks to improve multiple sections of the provided machine in attempt to allow it to provide reliable materials testing data.

2

3 The Need for Project As it is, this machine will not provide useful materials data. The ADMET test rig is built for a variety of material fatigue testing scenarios. However, when a user initiates a standard cyclical fatigue life test, the machine does not stay within the tolerances. This is because there is a certain amount of force amplitude undershoot or overshoot especially when the frequency is increased. The reason this is such a problem for the device is that without a reliable magnitude of force output any testing data would be worthless. The reported fatigue life of any sample tested using this machine most likely underwent too much or too little force. The ultimate result of this faulty data would be unsafe designs or projects that are more costly than need be because they are overdesigned. The Design Project Objectives and Requirements The new controller design must Design Objectives output a force load that matches The objective of our final design is to match the output force to the the input of the user. desired input of the user. The analysis of the mechanics of the machine needs to be discovered in order to understand how the thrust rod is controlled. The newly designed code needs to correctly penalize the system measured from the load cell. Design Requirements The controller design needs to output a force that is within 0.5% of the desired input from the user. Design Concepts considered The project was broken down Stiffness Dependent Algorithms into different aspects all which Another possible solution to the problem observed is to create needed to be addressed in order multiple algorithms for materials that have different stiffness values. to properly solve the project. This would factor in the numerous errors that have been reported as the materials are constantly changing for the same test. Multiple algorithms would have different constants based on the materials reaction in the S-N tests. Each algorithm would be specific for a certain tolerance of material stiffness. Structural Integrity Another possible solution would be to attack this problem from a mechanical sense. Originally the linear actuator was attached to a basic

4 aluminum structure which may not offer enough structural integrity during the tests. The actuator thrust rod gains momentum as it performs the sine waveform test. This momentum can cause flex within the aluminum structure causing the sample to be over or under compressed during the test. Also an ample amount of vibration can accrue within the structure during a high frequency sine waveform test. This may also cause overshoot or undershoot depending on how the structure responds to the vibration. Both the structures strength and stability of design and material are being called into question. Recommended Design Concept A simulation of the machine that Design Description can be easily manipulated and Split into hardware and software: As the end goal is to reduce or sent to a computer engineer for eliminate error in sinusoidal force application in a fully-constructed programming is recommended. system, the design concept takes the form of a series of modifications to aforementioned system. The first is simple yet robust; a stabilizing base was found to significantly reduce (but not remove) operating errors. This base was conceived and constructed using basic binration and moment-of-inertia calculations. The second modification is software/controls based, and it takes the form of a Simulink/MATLAB model that accurately simulates the experimental system, to the point to which realistic transfer functions may be extracted, modified and tested. The end product of this model serves as a deliverable to Admet s C programmers, who may then readily upload it to their system for use. Analytical Investigations Hardware: In order to see the effects that the frame had on the system, we designed a more robust frame to compare with the aluminum frame. The new frame was designed in SolidWorks and fabricated by the Northeastern machine shop. The main backbone beam of each system was analyzed and it was found that the stiffness is comparable. However, the connection points between beams are much sturdier on the robust frame. The robust frame was mounted to a 240 lb cast concrete block to prevent the frame as a whole from moving. Software: The system is dynamically adjusted by a PID controller (ref. 5). Adjusting the gains and the setup of the functions

5 between the inputs and output of the controller can create better performance. A model of the system and its controls allows for performance investigation. Experimental Investigations Hardware: The robust frame noticeably improved the results; transforming output data that was scattered into a clear and concise line. Although it did not solve the overall problem, this will improve the accuracy of the machine and give the controller more realistic data to work with. Software: A model of the machine in Simulink with a sinusoidal input put the problem in visual terms and broke the controller down into simpler parts for analysis. The figure below is an example of one of these investigations. Figure Model of Controller The model allows for adjustment of the input, the control gains, and the transfer function. Key Advantages of Recommended Concept The hardware concept is necessary, and required. The key advantage of the software concept is the design and implementation within a virtual physical environment, for ready modification by future designers and users, without the difficulty of, e.g. debugging or re-interpreting. Financial Issues The hardware aspect of this project has been the source of the only monetary costs. The total amount spent on hardware has This project has had two separate fronts since its beginning. The problem has been attacked from a software front which has been of zero cost to the project besides the man hours that it took to analyze and breakdown the code.

6 been 150 dollars. The hardware front of the project had a higher cost in monetary value. One hundred and fifty dollars was spent on outside materials to build the sturdier frame design. The steel used was recycled scraps from previous capstone projects in order to keep the cost down and all hardware was found in the machine shop. All technical machining was done in house by John Doughty to suppress costs. Recommended Improvements Implementation of multiple PID A recommended programming redesign of the basic Quattro control loops in order to improve architecture would be to change code from C-Code to a newer system flexibility. programming language such as C-Sharp would greatly reduce the complexity of the existing code structure. Another suggestion for tuning the PID loop would be the implementation of multiple PID loops within one another. This would create a more robust algorithm that would more closely match a wider range of materials used in fatigue testing. The response of the system would be more closely adjusted in order to accurately modify the response of the system. Though this would further complicate the programming, as long as the basic control function is functional, building upon it could only prove beneficial.

7